PARTITION TESTING DOES NOT INSPIRE CONFIDENCE

Partition testing, in which a program's input domain is divided according to some rule and tests conducted within the subdomains, enjoys a good reputation. However, comparison between testing that observes subdomain boundaries and random sampling that ignores the partition gives the counterintuitive result that partitioning is of little value. In this paper we improve the negative results published about partition testing, and try to reconcile them with its intuitive value. Theoretical models allow us to study partition testing in the abstract, and to describe the circumstances under which it should perform well at failure detection. Partition testing is shown to be more valuable when the partitions are narrowly based on expected failures and there is a good chance that failures occur. For gaining confidence from successful tests, partition testing as usually practiced has little value.